Slashdot Mirror
After 60 Years, Tape Reinserts Itself
Lucas123 writes "While magnetic tape is about as boring as technology gets, it's still the cheapest storage medium and among the fastest in sequential reads and writes. And, with the release of LTO-6 with 8TB cartridges around the corner and the relatively new open linear tape file system (LTFS) being embraced by movie and television markets, tape is taking on a new life. It may even climb out of the dusty archives that cheap disk has relegated it to. 'Over the last two years, disk drives have gotten bigger, they've gone from 1TB to 3TB, but they haven't gotten faster. They're more like tape. Meanwhile, tape is going the other direction, it's getting faster,' said Mark Lemmons, CTO of Thought Equity Motion, a cloud storage service for the motion picture industry."
Once again, Reel-To-Reel computers are no longer anachronistic in 60's Sci-Fi shows.
Sure, it reinserts itself, but when it's finished does it take itself out, flip it to the other side, and then reinsert itself again?
I have twenty terabyte backups NIGHTLY. I am required to keep certain tables (files by another name) for seven years but fortunately not all of it has to be online. I have over twenty terabytes I have to have backed up each night and a specific number of these backups available both on and off site. I have copies of quarterly and yearly complete backups I have too keep.
Show me a disk solution that is even remotely affordable. Cheap disk, maybe if you don't have any real amount of data and are not legally bound to keep it.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
Full disclosure: I work as a professional backup/recovery sysadmin. I have been working with tape for over seven years. It's not dead; far from it. Now, if you want to argue that there are areas where it used to make sense, but doesn't any more, I'd completely agree with you. But consider one usage case as an example of how tape is still incredibly useful - this is taken from a company I did work for a couple of years ago.
You have multiple petabytes of data. At any given point in time, you need to be able to access a specific subset of that data. You can predict, ahead of time, 95+% of the data that you will need in (say) a week's time. The rest, you don't need to be able to access quickly, but you have to retain it, because it's expensive (if not impossible) to reproduce. Once you're done with the data, you might not need it again for a year or more, or you might need it again in a few days (possibly a few hours.)
So: you could have a very large disk array to store all that data. The hard drives are relatively cheap individually, but the support infrastructure to merge them into large arrays is expensive. The cost to keep them all online (electricity, cooling) is high. The probability of failure is relatively low, but the rebuild time if it happens can be high, depending on how things are structured. The marginal cost of adding more storage is relatively high: you have to get a new array, new disks, new fibre connections, hook it all up, hope you have enough power, ...
Or you could have a large tape library, with multiple high-capacity drives (LTO4 in the case of the customer I'm thinking of), and thousands (yes, I'm serious) of cartridges. The data is written to the tapes; ideally duplicated (this customer didn't do that; I reckon they were stupid); and then deleted from the hard disk staging area (only a couple of TB in size.) When a given piece of data is needed, it's read off the tapes in advance, written to disk, and then accessed from the disk. Once it's no longer needed, it's simply deleted off the disk (since it's already on the tape.) Marginal cost of adding more storage: how much does a single tape cartridge cost? (maybe a storage frame if the library's full; they aren't exactly cheap, but they do hold over 1300 cartridges each: over a petabyte in potential capacity with no extra electrical requirements [LTO4; double that for LTO5], bang, done.) Electrical and cooling requirements: significantly lower; you only have to worry about a couple of TB of disk space, and a few tape drives, plus the tape robot. Rebuild time: just copy it off the redundant copy if the tape's bad.
Is this sort of usage typical? No, not really. But it's certainly not abnormal, and this is the sort of case where tape whomps all over disk when you sit down and work through all the numbers (look at it generally, rather than thinking just about the specifics outlined; they illustrate the point, and aren't the entire point themselves.) Tape's also useful if you need to move large quantities of data offsite (backups, anybody?) and can't afford, or don't want, high capacity fibre out of your data centre to another remote location.
I agree that "capacity after compression" is pure marketing; I do my figuring based upon native capacity (800 GB for LTO4; 1.5 TB for LTO5; 5 TB for T10000C; 3.2 TB - we hope - for LTO6). But to say that tape is "way too expensive, way too small, and ... completely irrelevant" is to misunderstand the strengths and uses of tape. Like I said: look at all the numbers, not just the purchase cost per raw TB, and pick whatever's right for the application in question.
Oh, and the use case I outlined above? It's for a pay TV network. TV shows, movies, sporting events, concerts, documentaries. All purchased legally, but impossible to reproduce (sporting events), or expensive to re-procure (TV shows, movies) if they're lost. Think about it.